Dissociating the generation stage in implicit and explicit memory tests: Incidental production can differ from strategic access

An assumption of the generate/recognize model of direct and indirect memory is that the generation stage is identical on explicit and implicit tests. Two experiments were conducted to examine the generation stage by requiring subjects to write down every word-stem solution they could generate on either an implicit test, a cued-recall test, or a generate/recognize test. In Experiment 1, the subjects studied words and anagrams; target generation was not significantly different on the three tests. However, in Experiment 2, the subjects studied the target words with a context word, and saw either thesame ordifferent context with the test stem. Now the generation stages dissociated, such that the context manipulation had no significant effect on the implicit test, but on the cued-recall test, more targets were generated with the same context words than were generated with different context words. The results argue against the claim that dissociations between implicit and explicit tests are due only to the addition of recognition processes on the explicit test, because the generation processes themselves can be dissociated.     

Video context-dependent recall

In two experiments, we used an effective new method for experimentally manipulating local and global contexts to examine context-dependent recall. The method included video-recorded scenes of real environments, with target words superimposed over the scenes. In Experiment 1, we used a within-subjects manipulation of video contexts and compared the effects of reinstatement of a global context (15 words per context) with effects of less overloaded context cues (1 and 3 words per context) on recall. The size of the reinstatement effects in Experiment 1 show how potently video contexts can cue recall. A strong effect of cue overload was also found; reinstatement effects were smaller, but still quite robust, in the 15 words per context condition. The powerful reinstatement effect was replicated for local contexts in Experiment 2, which included a nocontexts-reinstated group, a control condition used to determine whether reinstatement of half of the cues caused biased output interference for uncued targets. The video context method is a potent way to investigate context-dependent memory.     

Environmental context and human memory

Five experiments examined the effects of environmental context on recall and recognition. In Experiment 1, variability of input environments produced higher free recall performance than unchanged input environments. Experiment 2 showed improvements in cued recall when storage and test contexts matched, using a paradigm that unconfounded the variables of context mismatching and context change. In Experiment 3, recall of categories and recall of words within a category were better for same-context than different-context recall. In Experiment 4, subjects given identical input conditions showed strong effects of environmental context when given a free recall test, yet showed no main effects of context on a recognition test. The absence of an environmental context effect on recognition was replicated in Experiment 5, using a cued recognition task to control the semantic encodings of test words. In the discussion of these experiments, environmental context is compared with other types of context, and an attempt is made to identify the memory processes influenced by environmental context.     

Exploring long-term modality effects: Vocalization leads to best retention

Five experiments employing an incidental learning procedure explored the effects of different input modalities on the long-term retention of word lists. In each of the first four experiments, mixed-mode presentations featuring three different modes of presentation were employed. In the baseline presentation mode in each experiment, subjects were to read words silently. In the two additional presentation modes, subjects were to vocalize or read and hear (Experiment 1), read and hear or mouth (Experiment 2), vocalize or write (Experiment 3), and vocalize or write without seeing the written word (Experiment 4). In Experiment 5, separate groups of subjects were presented with pure-mode lists that they were to read silently, write without seeing, write and see, mouth, hear, read and hear, or vocalize. The principal findings were that auditory presentation procedures led to best memory performance and that, of these, only vocalization was found to consistently enhance retention. These findings are conceptualized within a framework that proposes that both the temporal distinctiveness of auditory information and self-generated cues are employed in the process of retrieval.     

Representation in the mental lexicon: Implications for theories of the generation effect

Three experiments investigated the finding by McElroy and Slamecka (1982) that the “generation effect” (the retention advantage for self-produced over read items) is not obtained when artificial, meaningless nonwords are used as the to-be-remembered items. In Experiment 1, some subjects were asked to generate or read items that they thought were words, but, in fact, were not; no generation effect was found. In Experiment 2, subjects were taught definitions to experimenter-created items. Despite the fact that these subjects could readily retrieve each item’s assigned semantic properties, no generation effect was found. Finally, Experiment 3 examined the read]generate variable as a function of an item’s frequency of use in the language. Whereas medium- and high-frequency words produced large generation effects, no comparable effects were found for low-frequency words or nonwords. These results indicate that representation in the mental lexicon is a necessary, but not a sufficient, condition for the generation effect. Rather, it may be necessary to consider how related the generated item is to other potential retrieval cues in the memory system.     

Meaning and modality: influences of context, semantic memory organization, and perceptual predictability on picture processing

Using event-related potentials (ERPs), the authors investigated the influences of sentence context, semantic memory organization, and perceptual predictability on picture processing. Participants read pairs of highly or weakly constraining sentences that ended with (a) the expected item, (b) an unexpected item from the expected semantic category, or (c) an unexpected item from an unexpected category. Pictures were unfamiliar in Experiment 1 but preexposed in Experiment 2. ERPs to pictures reflected both contextual fit and memory organization, as do ERPs to words in the same contexts (K. D. Federmeier & M. Kutas, 1999). However, different response patterns were observed to pictures than to words. Some of these arose from perceptual predictability differences, whereas others seem to reflect true modality-based differences in semantic feature activation. Although words and pictures may share semantic memory, the authors' results show that semantic processing is not amodal.     

The generation effect: Further tests of the lexical activation hypothesis

Three experiments compared recognition memory for word versus nonword responses when they had been either read or generated using a rhyme rule and either a word or nonword stimulus. That is, either the wordshop or the nonwordthop was generated from either the wordchop or the nonwordphop. In Experiment 1, the lexicality of the stimulus and response terms was manipulated orthogonally between subjects; in Experiments 2 and 3, it was manipulated orthogonally within subjects. In Experiment 3, subjects also made a lexical (word-nonword) decision about each response term after it had been read or generated. In all three experiments, memory performance was better for generated than read responses. This generation effect occurred only if the response term was a word, regardless of whether the stimulus term was a word or a nonword. These results are discussed in terms of the roles that lexical memory and response unitization play in the generation effect.     

Conversation memory: the effects of speaker status on memory for the assertiveness of conversation remarks

We conducted three experiments to examine the effects of information about a speaker's status on memory for the assertiveness of his or her remarks. Subjects either read (Experiments 1 and 2) or listened to a conversation (Experiment 3) and were later tested for their memory of the target speaker's remarks with either a recognition (Experiment 1) or a recall procedure (Experiments 2 and 3). In all experiments the target speaker's ostensible status was manipulated. In Experiment 1, subjects who believed the speaker was high in status were less able later to distinguish between remarks from the conversation and assertive paraphrases of those remarks. This result was replicated in Experiment 2, but only when the status information was provided before subjects read the conversation and not when the information was provided after the conversation had been read. Experiment 2's results eliminate a reconstructive memory interpretation and suggest that information about a speaker's status affects the encoding of remarks. Experiment 3 examined this effect in a more ecologically representative context.     

Generation enhances semantic processing? The role of distinctiveness in the generation effect

The goal of the present study was to investigate the locus of the memory advantage for words that are generated according to a nonsemantic rule (letter transposition) over words that are presented intact (read words). In the first two experiments, a category instance generation task was used to test the possibility that the semantic features of generated words are more readily available than those of read words. This possibility was not supported. In Experiment 3, generation effects were found to depend on the level of meaningfulness of words in recall, but not in recognition. In Experiment 4, a between-list design eliminated the generation effect found in recall, but did not affect the generation effect in recognition. Taken together, these findings suggest that generating a target according to a letter transposition rule enhances the distinctiveness of the word along a nonsemantic dimension.     

Context effects of pictures and words in naming objects, reading words, and generating simple phrases

In five language production experiments it was examined which aspects of words are activated in memory by context pictures and words. Context pictures yielded Stroop-like and semantic effects on response times when participants generated gender-marked noun phrases in response to written words (Experiment 1A). However, pictures yielded no such effects when participants simply read aloud the noun phrases (Experiment 2). Moreover, pictures yielded a gender congruency effect in generating gender-marked noun phrases in response to the written words (Experiments 3A and 3B). These findings suggest that context pictures activate lemmas (i.e., representations of syntactic properties), which leads to effects only when lemmas are needed to generate a response (i.e., in Experiments 1A, 3A, and 3B, but not in Experiment 2). Context words yielded Stroop-like and semantic effects in picture naming (Experiment 1B). Moreover, words yielded Stroop-like but no semantic effects in reading nouns (Experiment 4) and in generating noun phrases (Experiment 5). These findings suggest that context words activate the lemmas and forms of their names, which leads to semantic effects when lemmas are required for responding (Experiment 1B) but not when only the forms are required (Experiment 4). WEAVER++ simulations of the results are presented.     

Context effects of pictures and words in naming objects,reading words,and generating simple phrases

In five language production experiments it was examined which aspects of words are activated in memory by context pictures and words. Context pictures yielded Stroop-like and semantic effects on response times when participants generated gender-marked noun phrases in response to written words (Experiment 1A). However, pictures yielded no such effects when participants simply read aloud the noun phrases (Experiment 2). Moreover, pictures yielded a gender congruency effect in generating gender-marked noun phrases in response to the written words (Experiments 3A and 3B). These findings suggest that context pictures activate lemmas (i.e., representations of syntactic properties), which leads to effects only when lemmas are needed to generate a response (i.e., in Experiments 1A, 3A, and 3B, but not in Experiment 2). Context words yielded Stroop-like and semantic effects in picture naming (Experiment 1B). Moreover, words yielded Stroop-like but no semantic effects in reading nouns (Experiment 4) and in generating noun phrases (Experiment 5). These findings suggest that context words activate the lemmas and forms of their names, which leads to semantic effects when lemmas are required for responding (Experiment 1B) but not when only the forms are required (Experiment 4). WEAVER++ simulations of the results are presented.     

Words translated in sentence contexts produce repetition priming in visual word comprehension and spoken word production

Previous research with words read in context at encoding showed little if any long-term repetition priming. In Experiment 1, 96 Spanish–English bilinguals translated words in isolation or in sentence contexts at encoding. At test, they translated words or named pictures corresponding to words produced at encoding and control words not previously presented. Repetition priming was reliable in all conditions, but priming effects were generally smaller for contextualized than for isolated words. Repetition priming in picture naming indicated priming from production in context. A componential analysis indicated priming from comprehension in context, but only in the less fluent language. Experiment 2 was a replication of Experiment 1 with auditory presentation of the words and sentences to be translated. Repetition priming was reliable in all conditions, but priming effects were again smaller for contextualized than for isolated words. Priming in picture naming indicated priming from production in context, but the componential analysis indicated no detectable priming for auditory comprehension. The results of the two experiments taken together suggest that repetition priming reflects the long-term learning that occurs with comprehension and production exposures to words in the context of natural language.     

Memory for internally generated words in Alzheimer-type dementia: breakdown in encoding and semantic memory

The "generation effect" is a phenomenon in which words that are generated by the subject are remembered better than words which are read. The present experiments examined this effect in patients with mild-to-moderate dementia of the Alzheimer type (DAT), healthy elderly adults, and young adults under a variety of different encoding and retrieval conditions. Experiment 1 employed an intentional learning task with multiple study/test trials using the same list of words. While both the young and elderly adults exhibited higher recall for internally generated words than read words, the DAT patients failed to demonstrate the effect even after repeated exposures to the same stimulus list. Experiment 2 replicated this same pattern of results using an incidental learning paradigm with both recall and recognition tests. Various explanations as to why the DAT patients failed to show the generation effect were discussed with particular emphasis placed on the role of semantic memory and encoding failure.     

The influence of lexical and conceptual constraints on reading mixed-language sentences: Evidence from eye fixations and naming times

In two experiments, we explored the degree to which sentence context effects operate at a lexical or conceptual level by examining the processing of mixed-language sentences by fluent Spanish-English bilinguals. In Experiment 1, subjects’ eye movements were monitored while they read English sentences in which sentence constraint, word frequency, and language of target word were manipulated. A frequency × constraint interaction was found when target words appeared in Spanish, but not in English. First fixation durations were longer for high-frequency Spanish words when these were embedded in high-constraint sentences than in low-constraint sentences. This result suggests that the conceptual restrictions produced by the sentence context were met, but that the lexical restrictions were not. The same result did not occur for low-frequency Spanish words, presumably because the slower access of low-frequency words provided more processing time for the resolution of this conflict. Similar results were found in Experiment 2 using rapid serial visual presentation when subjects named the target words aloud. It appears that sentence context effects are influenced by both semantic/conceptual and lexical information.     

Acquisition-test interactions between

The effects of changes in context on recognition memory were investigated in two experiments. In Experiment 1 target items were embedded in congruous or incongruous sentence frames and then tested in the original, new congruous, or new incongruous contexts. Experiment 2 included a third encoding condition; targets were also embedded in abstract definitional sentences. Both experiments revealed a reliable crossover interaction between the initial encoding condition and the mode of the changed context. Congruous items were recognized better in new congruous contexts, whereas incongruous items were recognized better in new incongruous contexts. Recognition of definitional items was equally impaired by both context changes. The results suggest the importance of the symmetry between acquisition and test encoding dimensions. These experiments also point out the need to distinguish between different ways in which the concept “integration” is used in current research.     

Capturing conceptual implicit memory: The time it takes to produce an association

Conceptual implicit memory is demonstrated when, in the absence of explicit retrieval, performance on a task requiring conceptual processing benefits more from prior conceptual encoding than from prior nonconceptual encoding. In the present study, we sought to provide an improved measure of conceptual implicit memory by minimizing contamination from explicit retrieval. On a modified word association test, participants free-associated to the actually studied items, with response time to produce any associate serving as the dependent measure. Experiment 1 varied whether words were read or generated at study and showed that generated words were associated to more quickly than were read words. Experiment 2 varied level of processing at study and showed that words processed semantically were associated to more quickly than were words processed nonsemantically. With modifications to ensure its implicit nature, a conceptual implicit test can be consistently affected by the same encoding manipulations as those that affect conceptual explicit memory tests.     

How generative drawing affects the learning process: An eye‐tracking analysis

Generative drawing is a learning strategy in which students draw illustrations while reading a text to depict the content of the lesson. In two experiments, students were asked to generate drawings as they read a scientific text or read the same text on influenza with author‐provided illustrations (Experiment 1) or to generate drawings or write verbal summaries as they read (Experiment 2). An examination of students' eye movements during learning showed that students who engaged in generative drawing displayed more rereadings of words, higher proportion of fixations on the important words, higher rate of transitions between words and workspace, and higher proportion of transitions between important words and workspace than students given a text lesson with author‐generated illustrations (Experiment 1) or students who were asked to write a summary (Experiment 2). These findings contribute new evidence to guide theories for explaining how generative drawing affects learning processes.     

Don’t Shoot the Messnger: Memory for Misspellings in Context

Misspellings in sentences are usually easy to understand by readers due to top-down influences. Although top-down processing allows for fluent reading of misspelled items, the nature of their representations in memory is not known. If representations of misspellings are distinct from representations of correctly spelled words, their influence should be seen in later recognition decisions. In this set of experiments, participants read words and misspellings embedded in sentences and were later given a recognition test. The sentences contained semantically biased or neutral contexts. In Experiment 1, misspellings were created by removing a single letter (e.g., drveway). In Experiment 2, the recognition items probes were presented in uppercase letters (e.g., DRVEWAY) to reduce the visual similarity between study and test items. In Experiment 3, the misspellings were created by substituting visually similar letters (e.g., driweway). In contrast to the previous experiments, in Experiment 4, participants were explicitly told about the memory test to see how response strategies affect performance. Overall, the results indicate that people retain surface feature information about misspellings which seem to inform their memory judgments, and that the processing of this information cannot be strategically controlled.     

Activation of semantic and phonological codes during reading

Three experiments used a new paradigm to examine word-coding processes during reading. Subjects read text passages and occasionally responded to lexical-decision probes. Experiment 1 focused on semantic and surface codes. The activation of semantic codes appeared to increase over time, whereas surface codes did not. In addition, subjects who were instructed to remember the exact wording of the passages showed stronger activation of surface codes than did subjects who read for comprehension. Experiments 2 and 3 explored the role of phonological codes. Experiment 2 showed that phonological codes were activated by specific words in a passage. In contrast, Experiment 3 found no evidence that phonological codes were activated by the more general passage context. Taken together, the experiments suggest some of the roles semantic and phonological codes may play during reading.     

Eye movements of highly skilled and average readers: Differential effects of frequency and predictability

This study offers a glimpse of the moment-by-moment processes used by highly skilled and average readers during silent reading. The eye movements of adult readers were monitored while they silently read sentences. Fixation durations and the spatial-temporal patterns of eye movements were examined to see whether the two groups of readers exhibited differential effects of frequency and/or predictability. In Experiment 1, high- and low-frequency target words were embedded in nonconstraining sentence contexts. In Experiment 2, the same participants read high- and low-frequency target words that were either predictable or unpredictable, embedded in highly constraining sentence contexts. Results indicated that when target words appeared in highly constraining sentence contexts, the average readers showed different effects of frequency and predictability from those shown in the highly skilled readers. It appears that reading skill can interact with predictability to affect the word recognition processes used during silent reading.